CDC STAR-100
ulm...@fafner.zdv.uni-mainz.de
Currently I am (quite desparately) looking for information regarding the
Control Data STAR-100 machine for my Ph.D. thesis which is about an extended
vector computing model. As far as I knwo the STAR-100 was designed with APL in
mind, is this right? Is there anybody out there who has by chance an
architecture manual of this early vector processor designed by Thornton? Or a
sample assembly language program from the old days? I will happily pay all
shipping and photocopying expenses! :-)
Why is it so difficult to find more about this machine? It is quite easy to
read about the 6600/7600 and even 8600 series and all CRAY-x machines, but the
STAR seems like a fata morgana to me. All I could find are some snippets from
"Computer Arithmetic / Principles, Architecture and Design" by Kai Hwang and
some stories about the machine but no article, no text book I searched had a
real code example or even an instruction list or something like that.
Thank you very much in advance for your help! Best regards, sincerely yours -
Bernd Ulmann. :-)
Douglas A. Gwyn
> Currently I am (quite desparately) looking for information regarding
I think Gordon Bell has something about it in his book; check out his
Web site. Also, the Charles Babbage Institute has documents, but you
may have difficulty in getting copies in time if you don't go there in
person. (I assume you can use a Web search engine to find these.)
> Why is it so difficult to find more about this machine?
(1) It predates on-line documentation.
(2) Few people had experience with it.
(3) Most people who had experience with it have other things to do.
I.e. same reason why it is hard to find information about *any* old
computers, other than a few very popular ones.
I'm still desperately seeking *any* software for the CDC 1700/
System 17/Cyber-18 series. 9-track magnetic tape, paper tape, or
hardcopy listings preferred.
Rob Peglar
mind,
but the effect of the architecture was such that many people considered it
to be the world's best APL machine. It certainly was designed with its
intended customer base, mostly US-based 'black' (secret) organizations and
national
labs, in mind, and the computing challenges they had in those days.
There were very few STAR machines made and delivered, in comparison with
their RISC brethern, the 6000 and 7000 series. The STAR is considered by
some
(me included) to be the CISC-iest machine ever designed. When you try to
deal with STring ARrays, after all, you get into some interesting microcode.
There is very little existing documentation on the STAR that is available.
Most
of it is hoarded by those who worked on the machine, and as Doug Gwyn put
it,
they have better things to do these days. But, a careful perusal of ETA-10
and
Cyber 20x architecture (esp. the 203) will give clues into their
predecessor, the STAR-100
and the original STAR-1A and -1B.
Rob
<ulm...@fafner.zdv.uni-mainz.de> wrote in message
news:9d74el$9i$1...@bambi.zdv.Uni-Mainz.DE...
Peter Schow
"A Safari through the Control Data STAR-100 with Gun and Camera"
by Neil Lincoln
which appeared in:
AFIPS Conference Proceedings, Vol 47, NCC, June 1978
This is an obscure but seminal (I think) paper that provides an honest
look
at what worked and what didn't, in the development of the STAR-100
during
that era (1964-1969). For example, he talks about the "goofy" ideas
that were
floating around inside their team, that would have been shot down in
hindsight,
such as:
- Ridiculous schedules for design, development, and delivery
- Accepting penalty clauses in contracts because someone believed the
ridiculous schedules
- Everyone would learn how to vectorize because (a) it's natural, (b)
it's
exciting and rewarding, and (c) they would quickly realize that there
was no alternative to future supercomputing
and others
On the APL influence, Neil Lincoln definitely mentions here that the
early STAR
architects in 1964 had Iverson's book open to Chapter 1. Not that the
APL
operations were directly copied, but it's clear the influence was there
early on. Later on in the late 60s, (see "Where Should the Dollars Go -
Iverson or Delivery" section), he talks about the conflict between
adding more APL operators vs. real-estate, reliability, performance, and
schedule pressures.
---
Peter Schow
Eugene Miya
Rob Peglar <peg...@earthlink.net> wrote:
>It certainly was designed with its
>intended customer base, mostly US-based 'black' (secret) organizations and
>national labs, in mind, and the computing challenges they had in those days.
I'm not certain that the Fort ever took one. Or the CIA. Or other
black organizations.
I know half a dozen folk at LLNL who worked on it.
It's generally regarded as "bad blood." There was also one at LaRC
which was upgraded to a Cyber 203. You left those off.
The 100 was apparently a very sore point and people got axed, because
of it (the Cray-1 at LANL also had some bad blood with Deimos, but those
people survived to inflict revenge).
>There were very few STAR machines made and delivered, in comparison with
>their RISC brethern, the 6000 and 7000 series. The STAR is considered by
>some (me included) to be the CISC-iest machine ever designed. When you try to
>deal with STring ARrays, after all, you get into some interesting microcode.
>
>There is very little existing documentation on the STAR that is available.
>Most of it is hoarded by those who worked on the machine, and as Doug Gwyn
>put it,
>they have better things to do these days. But, a careful perusal of ETA-10
>and Cyber 20x architecture (esp. the 203) will give clues into their
>predecessor, the STAR-100 and the original STAR-1A and -1B.
I can find no pieces of substantial size for the Museum.
We have one photo.
Doug noted:
>I think Gordon Bell has something about it in his book; check out his
>Web site. Also, the Charles Babbage Institute has documents, but you
>may have difficulty in getting copies in time if you don't go there in
>person. (I assume you can use a Web search engine to find these.)
I don't believe that it's in Sieworek, Bell, and Newell.
www.research.microsoft.com and you can reach ~gbell's web page.
It's not a Seymour design, do not bother looking down that link for
the 100 or the 205.
>> Why is it so difficult to find more about this machine?
>(1) It predates on-line documentation.
>(2) Few people had experience with it.
>(3) Most people who had experience with it have other things to do.
>I.e. same reason why it is hard to find information about *any* old
>computers, other than a few very popular ones.
No, it goes beyond that. People actively wanted to quash the existence
of the machine. The AEC, later ERDA, was not particularly happy with the
amount of money poured into it.
There were some apparently some serious psychologist cases after the
experience.
Eugene Miya
Peter Schow <Peter...@Sun.MOC> wrote:
>My favorite STAR-100 paper is:
> "A Safari through the Control Data STAR-100 with Gun and Camera"
> by Neil Lincoln
>which appeared in:
> AFIPS Conference Proceedings, Vol 47, NCC, June 1978
>
>This is an obscure but seminal (I think) paper that provides an honest
>look at what worked and what didn't,
I also endorse this paper.
Additionally, I endorse Neil's other papers:
%A N. R. Lincoln
%T Great Gigaflops and Giddy Guarantees
%B Supercomputing
%E N. Metropolis
%E D. H. Sharp
%E W. J. Worlton
%E K. R. Ames
%S Los Alamos Series on Basic and Applied Sciences
%V 7
%I University of California Press
%C Berkeley, CA
%D 1986
%P 16-24
%K trade/popular/business press, industry references, ETA,
%X An amusing talk! Not much technical content, but amusing.
%A Neil R. Lincoln
%T Technology and Design Tradeoffs in the Creation of a Modern Supercomputer
%B Supercomputers, Class VI Systems, Hardware and Software
%E Sidney Fernbach
%I North-Holland
%C Amsterdam
%D 1986
%P 83-112
%K Cyber 205, vector processing,
%K trade/popular/business press, industry references,
%X See the original IEEE TOC article, C-31, no 5, may '82, pages
349-362.
%A Neil R. Lincoln
%T Supercomputers = Colossal Computations + Enormous Expectations +
Renowned Risk
%J IEEE Computer
%V 16
%N 5
%D May 1983
%P 38-47
%K CDC Star-100, Cyber-203, Cyber-205,
%X Another paper which is a retrospective (rather apologetic) on the
CDC Star-100, Cyber 203/205 development.
%A Neil R. Lincoln
%T Technology and Design Tradeoffs in the Creation of a Modern Supercomputer
%J IEEE Transactions on Computers
%V C-31
%N 5
%D May 1982
%P 349-362
%K architecture, parallelism, pipeline, supercomputer, technology,
vector processor,
frecommended, special issue on supersystems
%X Architectural survey of the STAR-100, Cyber 203/205 line
Also printed in Fernbach's "Supercomputers, Class VI Systems..."
book, North-Holland, 1986.
Also reprinted in the text compiled by Kai Hwang:
"Supercomputers: Design and Application," IEEE, 1984, pp. 32-45.
See also the "not as Fun" article.
%A Neil R. Lincoln
%T "Its really not as much fun building a supercomputer as
it is simply inventing one"
%B High Speed Computer and Algorithm Organization
%E D. J. Kuck
%E D. H. Laurie
%E A. H. Sameh
%I Academic Press
%C New York
%D 1977
%P 3-11
%K frecommended, Cyber 205,
Ginsberg biblio:
%X Delightful, entertaining, yet very informative about the
real problems of actually building a supercomputer.
The story has not changed much in the last 10 years.
A "must" reading for anyone interested in computer architecture.
Amos Omondi
and...
%A Lloyd Thorndyke
%T The Demise of ETA Systems
%B Frontiers of supercomputing II: a national reassessment
%E Karyn R. Ames
%E Alan Brenner
%S Los Alamos series in basic and applied sciences
%V 12
%I University of California Press
%I Berkeley
%D 1994
%P 489-496
%K experience and lessons learned,
%X Good commentary on non-technical management.
Paragraph 1: we started with too many people. Division of Cyber 205 and
ETA lines were a problem (credit and punishment).
LN2: "We could replace a processor in a matter of hours."
[Cray-2 CPUs could be replaced in 20 minutes.]
Comments about software 1/2 page, hardware: 2 pages.
Industry observations: vertical integration: chip makers not responsive.
Gregory Travis
Here's some historical information which I've not yet seen posted in this
thread.
STAR-100 was designed in the advanced systems laboratory by J.E. Thornton
and Neil Lincoln. Thornton was, of course, Cray's protege and the author
of "Design of a Computer: The CDC 6600." Thornton, Vice President for
Advanced Systems, left CDC in 1973 just as the STAR-100 sh*t must have
really started to hit the fan. I don't know if he was pushed out,
sensed that he had to get out, or if there were other factors which
led to his decision to leave.
Someone else may correct me but I believe that it was CDC's decision
to drop Cray's 8600 and fund only Thornton's STAR-100 which was
the causal event which led to Cray's leaving CDC in 1972. IIRC, it
was originally decided to develop both machines but budget considerations
later made that impossible. CDC decided to keep the STAR-100 program
and kill the 8600. Interestingly, although this was the first time
that a protege upped the master in the corporation's eye, it wasn't
the last. Cray would have to live through the experience again when
Cray Research chose Steve Chen's upgraded Cray-1 over Cray's Cray-3
design and Cray, once again, had to leave his company. Even more interesting
is the change of circumstances. In the first case, CDC decided to
abandon a relatively conservative upgrade to their existing line
(i.e. the 7600->8600) in order to concentrate on a machine that
was a radical departure from the tried and true. In the
second case, Cray Research decided to abandon the new architecture
(the Cray-3) in order to concentrate on incremental improvements
to their existing architectures. Cray himself was on the opposite
side of each situation!
Anyway, the STAR-100 was a radical departure from the 6000
series. It was, as I recall, a memory/memory machine and
had no internal registers. Cray must have shuddered at the
complexity of the machine.
In 1979 the STAR-100 hardware was redone. I don't recall
the changes but the gist was faster parts and thus a
faster machine. This machine was called the
Cyber 203E. In 1982 the machine was again upgraded
with faster parts and became the Cyber 205. That design
eventually morphed into the ETA-10 machine and then things
fell completely apart.
greg
Jgzabol
Travis) schreibt:
>
>Someone else may correct me but I believe that it was CDC's decision
>to drop Cray's 8600 and fund only Thornton's STAR-100 which was
>the causal event which led to Cray's leaving CDC in 1972. IIRC, it
>was originally decided to develop both machines but budget considerations
>later made that impossible. CDC decided to keep the STAR-100 program
>and kill the 8600. Interestingly, although this was the first time
>that a protege upped the master in the corporation's eye, it wasn't
>the last. Cray would have to live through the experience again when
>Cray Research chose Steve Chen's upgraded Cray-1 over Cray's Cray-3
>design and Cray, once again, had to leave his company. Even more interesting
>is the change of circumstances. In the first case, CDC decided to
>abandon a relatively conservative upgrade to their existing line
>(i.e. the 7600->8600) in order to concentrate on a machine that
>was a radical departure from the tried and true. In the
>second case, Cray Research decided to abandon the new architecture
>(the Cray-3) in order to concentrate on incremental improvements
>to their existing architectures. Cray himself was on the opposite
>side of each situation!
>
>
The 8600 story is told quite differently in the book "The Supermen" by Murray.
There is it said that the 8600 was _NOT_ a more or less straightforward
upgrade of the 7600, but a radically different parallel-processing design.
It is also said that the decision to stay with discrete parts lead to having
too
many of them, so that for simple statistical reason the machine never
could work reliably. It is also said that Cray himself realized that he had to
abandon the 8600. Some modules are at display at the Chippewa Falls
Museum.
John G. Zabolitzky
Douglas A. Gwyn
> Some modules are at display at the Chippewa Falls Museum.
Which museum is that? Does it have a Web page?
Jgzabol
schreibt:
>
>Jgzabol wrote:
>> Some modules are at display at the Chippewa Falls Museum.
>
>Which museum is that? Does it have a Web page?
>
>
http://www.execpc.com/~cfmit/cfmit.htm
Douglas H. Quebbeman
news:20010611142714...@nso-ch.aol.com...
Unfortunately, they seem to have been having a problem for a while...
several pictures can't be found, and the link to the virtual Cray Museum
at the Charles Babbage Institute has also gone dead. CBI may or may
not be aware...
-dq
Shannon
>It's generally regarded as "bad blood." There was also one at LaRC
>which was upgraded to a Cyber 203. You left those off.
Where was the Star at Langely? The 203? I tried to find out about
machines like that when I was in college, and was either told "I don't
know" or emphatically that LaRC never had a Cyber or related machine.
Note, I'm not saying there wasn't; I didn't have much faith in the
answers at the time.
There were several Cray there over the years, though I only got to see
one of them, and it was in the 90s at Lockheed.
Someone once told me that the US Navy had a very large Cyber 205 site
somewhere in this area (Tidewater, Virginia). In an 80s magazine there
was a question sent in by a reader asking "What is the largest computer
system in the world?" and the answer given was a US Navy Cyber 205,
and I think they said it was in Virginia. The article said it had a
very large mass storage system supporting it, but it's been so long I
forget the other details. In college I tried to find the article again
because I was doing research on military systems, and queries to the
Navy produced no results (they never answered, not even to deny my
request for information).
--
sha...@widomaker.com _________________________________________________
______________________/ armchairrocketscientistgraffitiexenstentialist
"And in billows of might swell the Saxons before her,-- Unite, oh
unite! Or the billows burst o'er her!" -- Downfall of the Gael
Eugene Miya
Shannon <sha...@widomaker.com> wrote:
>In article <3af88ee2$1...@news.ucsc.edu>, Eugene Miya <eug...@cse.ucsc.edu> wrote:
>
>Where was the Star at Langely? The 203? I tried to find out about
>machines like that when I was in college, and was either told "I don't
>know" or emphatically that LaRC never had a Cyber or related machine.
>Note, I'm not saying there wasn't; I didn't have much faith in the
>answers at the time.
I never saw it.
I've only been to LaRC once, but the people I know who used it were the
LaRC people and the ICASE people like the Voigts. Funny you should
mention this, I'm looking at a copy of Myron Ginsberg's bibliography of
about that time, and I am adding Myron's annotations to my parallel
computing biblio.
>There were several Cray there over the years, though I only got to see
>one of them, and it was in the 90s at Lockheed.
Oh yes, they had a 2 (had some involvement with OS selection [we called
it a war, Rollwagon wrote his "dragged us kicking and screaming into the
Unix era" letter of thank you)]. The only machine I ever really saw
there was the FLEX (I have a photo, I need to collect one for The Museum).
And it was the replacement for the FEM.
>Someone once told me that the US Navy had a very large Cyber 205 site
>somewhere in this area (Tidewater, Virginia). In an 80s magazine there
>was a question sent in by a reader asking "What is the largest computer
>system in the world?" and the answer given was a US Navy Cyber 205,
>and I think they said it was in Virginia. The article said it had a
>very large mass storage system supporting it, but it's been so long I
>forget the other details. In college I tried to find the article again
>because I was doing research on military systems, and queries to the
>Navy produced no results (they never answered, not even to deny my
>request for information).
Oh, it's hard to say. We were told we had the only 4 pipe 205 in the world.
There were likely others like at FNOMC (whatever spelling).
I think in the end it's all moot. One of these days I will collect the
list of old 205 host sites, but right now I'm just trying to get the list
of Cray-2 sites release into the public domain. I got us at least 2 ETA
cryostats with Neil's help.
Questions of sizes will likely never fully be known. I've seen part of
"The Fort's" collection as well as LLNL and LANL, and those are big
honking facilities.
I would look for 203/205 parts to try to preserve it's history better,
but there are far more important machines for me to seek. I have to
leave it to my friends who are ex-CDC to come forward, organize
themselves and get their story out. I have sympathy, but the Museum
has the 6600 and 7600 and tons of documents. But it's not ever clear
what will happen with them except to collect dust.
It's as bad as pulling teeth to try to get the LLNL guys to talk about
STAR-100 stories.
I've only got so much time to help the Museum.
Shannon Hendrix
> I never saw it.
> I've only been to LaRC once, but the people I know who used it were the
> LaRC people and the ICASE people like the Voigts. Funny you should
> mention this, I'm looking at a copy of Myron Ginsberg's bibliography of
> about that time, and I am adding Myron's annotations to my parallel
> computing biblio.
I've been in ICASE before, hoping to get hired as an administrator.
Interesting place, and interesting work. I got the grand tour, but
the only "super computers" they had there were stacks of networked
Sun workstations. I can't remember what they said they were connected
to there.
Don't know who the Voigts were. I met Manny Salas, the director, and
head of computer science.
> Oh yes, they had a 2 (had some involvement with OS selection [we called
> it a war, Rollwagon wrote his "dragged us kicking and screaming into the
> Unix era" letter of thank you)]. The only machine I ever really saw
> there was the FLEX (I have a photo, I need to collect one for The Museum).
> And it was the replacement for the FEM.
At the Virginia Air and Space Museum in downtown Hampton, there is a
Cray on display. I think it is a one, but I can't really remember. I'm
not sure where they got it. It has the CPU, at least one big nitrogen
system (with the cool bubbling plexiglass), and another cabinet I can't
remember. I should go look again. I can't believe how Cray not only
built a fast machine, but made it look so damn cool too. IBM would
have hidden the coolant so you couldn't see it.
It seems to be in very good condition.
> Oh, it's hard to say. We were told we had the only 4 pipe 205 in the world.
> There were likely others like at FNOMC (whatever spelling).
The magazine claimed this 205 was supported by something like 271 hard
drives in a rather large room just for that. I can't remember any other
details, but it mentioned how much memory it had and a few other things.
Something I guess I should have saved.
> I think in the end it's all moot. One of these days I will collect the
> list of old 205 host sites, but right now I'm just trying to get the list
> of Cray-2 sites release into the public domain. I got us at least 2 ETA
> cryostats with Neil's help.
The Cray I mentioned in the museum had an explanation of it's history
if I recall correctly. You interested in that?
> Questions of sizes will likely never fully be known. I've seen part of
> "The Fort's" collection as well as LLNL and LANL, and those are big
> honking facilities.
What is "The Fort". I've been meaning to ask.
> I would look for 203/205 parts to try to preserve it's history better,
> but there are far more important machines for me to seek.
They sold something from one on eBay some time back. Now and then you
seem people selling pieces of various machines. You will also see
the occasional manual as well.
> I have to leave it to my friends who are ex-CDC to come forward,
> organize themselves and get their story out. I have sympathy, but
> the Museum has the 6600 and 7600 and tons of documents. But it's
> not ever clear what will happen with them except to collect dust.
Well, that's better than them being sent to scrap at least. As much as
I like preservation, I have to admit it's very difficult to keep one on
display. It costs money, and the general public won't pay to see things
it has no interest in.
> It's as bad as pulling teeth to try to get the LLNL guys to talk about
> STAR-100 stories.
Interesting. The Multics and PDP fan-site are FULL of old stories.
> I've only got so much time to help the Museum.
Yeah, I need to be doing everything but reading Usenet and anything
else it might lead to... :)
--
"Star Wars Moral Number 17: Teddy bears are dangerous in herds."
Rob Peglar
"Eugene Miya" <eug...@cse.ucsc.edu> wrote in message
news:3b4f6a76$1...@news.ucsc.edu...
>
> Oh, it's hard to say. We were told we had the only 4 pipe 205 in the
world.
Outside ARHOPS, yes. 501 was a 4-pipe machine. 505 might have been, also;
IIRC, the commercial folks tried to sell 4-pipe machine time. Not very
successful.
> There were likely others like at FNOMC (whatever spelling).
FNOC never had a 4-pipe 205.
> I think in the end it's all moot. One of these days I will collect the
> list of old 205 host sites, but right now I'm just trying to get the list
> of Cray-2 sites release into the public domain. I got us at least 2 ETA
> cryostats with Neil's help.
I don't think you'll ever get a complete list of 205 sites, unless Neil has
it stashed away somewhere.
> I would look for 203/205 parts to try to preserve it's history better,
> but there are far more important machines for me to seek. I have to
> leave it to my friends who are ex-CDC to come forward, organize
> themselves and get their story out. I have sympathy, but the Museum
> has the 6600 and 7600 and tons of documents. But it's not ever clear
> what will happen with them except to collect dust.
>
> It's as bad as pulling teeth to try to get the LLNL guys to talk about
> STAR-100 stories.
They're still mad that we didn't use Livermore Red/LTSS. Well, we used
parts of it, anyway.
I do have a STAR-100 part if you really want it.
Rob
Shannon Hendrix
Shannon Hendrix <sha...@widomaker.com> wrote:
> Don't know who the Voigts were. I met Manny Salas, the director, and
> head of computer science.
Oops... meant to say Thomas E., head of computer science.
--
"An Irishman is never drunk as long as he can hold onto one | | |
blade of grass and not fall off the face of the earth." | | |
________________________________________________________________ / | \
s h a n n o n @ w i d o m a k e r . c o m _/ | \_
Eugene Miya
Shannon Hendrix <sha...@widomaker.com> wrote:
>In article <9ios54$k5m$1...@daydream.shannon.net>,
>Shannon Hendrix <sha...@widomaker.com> wrote:
>> Don't know who the Voigts were. I met Manny Salas, the director, and
>> head of computer science.
>
>Oops... meant to say Thomas E., head of computer science.
Actually, I've heard it said that "We don't do computer science in NASA."
I've never heard of Salas, but I know Bob and his wife.
%A James M. Ortega
%A Robert G. Voigt
%A Charles H. Romine
%T A Bibliography on Parallel and Vector Numerical Algorithms
%R Interim Report 6
%I ICASE
%C Hampton, VA
%D December 1988
%K numerical methods for parallel computation,
parallel computer architecture, scientific computing,
technical report, rbibliography,
%X A good, but large paper survey (largely the Ricase2) file on this work.
It does not cite this bibliography (Miya's).
%A Jack J. Dongarra
%A Paul Messina
%A Danny C. Sorensen
%A Robert G. Voigt, eds.
%T Parallel Processing for Scientific Computing
%I SIAM
%C Phil., PA
%D 1990
%K book, text,
%X Proc. 4th SIAM conference Proc. Applied Math 44.
Matrix computations, numerical method, differential equations,
massively parallel computing, performance and tools.
%A Jack Dongarra
%A Ken Kennedy
%A Paul Messina
%A Danny C. Sorenson
%A Robert G. Voigt, eds
%J Proceedings of the Fifth SIAM Conference on Parallel Processing for
Scientific Computing
%I SIAM
%C Houston
%D 1992
%K book, text,
%A Piyush Mehrotra
%A Joel Saltz
%A Robert Voigt, eds.
%T Unstructured Scientific Computation on Scalable Multiprocessors
%I MIT Press
%C Cambridge, MA
%D 1992
%K book, text,
%A R. Voigt
%A D. Gottlieb
%A M. Hussaini, eds.
%T Spectral Methods for Partial Differential Equations
%I SIAM
%C Philadelphia
%D 1984
And numerous other pubs.
fraho...@gmail.com
http://www.chilton-computing.org.uk/acl/literature/minutes/srcaccmtgs/star100.htm
Daiyu Hurst
> There is some information at:
>
> http://www.chilton-computing.org.uk/acl/literature/minutes/srcaccmtgs/star100.htm